Pér
Glossary¶
These are the terms this documentation uses, and what each one means in Pér.
Several of them are everyday English words with a specific meaning in Pér. For example, A plan is a particular object with a particular lifecycle, not just an intention. A memory is something you asked Pér to keep, not everything it has seen.
- Activity log
The record of what Pér did in your tenant: the actions it took and the recommendations it made. Amperity keeps its own, separate activity logs covering everything that happens across the platform.
- AI agent for customer data
The kind of product Pér is: an agent that works on your customer data, answering questions about it and — with your approval — acting on it in Amperity.
- Allow Pér access
The grant that lets one person into Pér when a tenant uses managed access. Given and removed in Amperity, on the Users page.
- apply mode
How a memory that Pér proposes gets saved. By default Pér asks every time. You can choose to let your personal memories save without asking; a memory shared with your tenant always asks, under either setting.
- Approve & run all
Approving a whole plan at once and letting it run itself. You approve the plan, not each write inside it: Pér re-checks at every step whether it may still go on, stops at any step that needs a person, and records which steps it approved on your behalf.
- artifact
Something a session leaves behind: a report Pér wrote, or a file you gave it. You can come back to an artifact later, export it to PDF, and share it so that everyone in your tenant can open it.
- Block user
The Amperity control that revokes all of a person’s access to a tenant, overriding any permission they hold directly or inherit. Pér has no revocation of its own — blocking happens in Amperity.
- company context
The business priorities, definitions and measures you want Pér to work from. Company context goes into every session. Alongside it, Pér reads the context documents and the AI Assistant system prompt your tenant has set up in Amperity; it reads those, and does not replace them.
- customer decision loop
The cycle Pér is built around: Understand → Recommend → Approve → Act → Learn.
- Guideline
One of the two enforcement levels of a memory. A guideline shapes what Pér does without binding it.
- Managed access
One of the two ways a tenant admits people to Pér: only people granted access individually can enter. Set in Amperity, on the Users page, by a User Administrator.
- MCP tools
The Amperity tools Pér calls to read your data and to change things on your behalf. Pér reads before it writes, every change becomes something you approve, and some tools are withheld from Pér entirely.
- memory
Something you have told Pér to remember between sessions — a preference, a rule, a fact or a correction. Each memory is either personal to you or shared with everyone in your tenant, and is either a Guideline or Must follow. Memories can be saved from a conversation or written by hand, and can be archived and restored. They do not expire.
- Must follow
One of the two enforcement levels of a memory. Pér reads must-follow memories first and treats them as rules it must not break.
- Open access
One of the two ways a tenant admits people to Pér: anyone authorized for the tenant can enter. Either way a tenant is set up, each person’s existing Amperity permissions still govern what they are able to do once inside.
- Pér
Amperity’s AI agent for customer data, and the name used throughout this documentation.
- plan
A titled, reviewable list of Amperity writes that you approve before any of them runs. A plan has steps, flags anything it considers risky, and collects what it produces. Plans are authored two ways: by acting on a recommendation, or by asking for one in conversation. A step that fails shows what went wrong and can be run again.
Reverting a plan abandons one that has not started yet, which frees its recommendation to be acted on again. It is not a way to undo work that has already run.
- Portfolio
Where Pér gathers the work it thinks is worth your attention: the recommendations it has made, and the plans already under way.
- recommendation
Something Pér proposes doing, with the argument attached: the claims it rests on, the numbers behind them, where in your data they came from, a confidence grade and the reasoning for that grade, and what Pér could not establish. Acting on a recommendation authors a plan. Once you have acted on one, Pér stops offering it.
- skill
A packaged piece of work you can start by name rather than describing from scratch. Which skills are available depends on your tenant.
- step
A single approvable unit inside a plan: one concrete change to your Amperity tenant. Steps are what you approve, either one at a time or all at once.
- trusted customer context
What Pér works from: your identity-resolved customer data, the history in it, the rules you have given Pér to work inside, and the predictive intelligence available in your tenant — all of it delivered by the Amperity platform. See Trusted customer context.
- User Administrator
The Amperity policy held by the people who manage Pér access: switching a tenant between open and managed access, and granting or removing access for individuals.
- write confirmation
What stands between Pér proposing a change to Amperity and that change happening. Nothing runs until a person works through it.